Image Classification on ImageNet-1k 5% annotated (train)
36.01AccuracyOurs-Joint
Evaluation Results
| Method | Links | |
|---|---|---|
| Ours-JointTraining strategy=Joint (without weight sharing)2021.12 | 36.01 | |
| ResNet50Backbone=ResNet-502021.12 | 35.43 | |
| Agent-BModel=Agent-B2021.12 | 35.28 | |
| Ours-SharedTraining strategy=Shared (with weight sharing)2021.12 | 33.06 | |
| ViT-B-SAMBackbone=ViT-B, Optimizer=SAM2021.12 | 16.67 | |
| ViT-BBackbone=ViT-B2021.12 | 16.6 | |
| ViT-B-SparseBackbone=ViT-B, Variant=Sparse2021.12 | 10.39 |